V. F. Solomatin

The method of computer modelling receives estimations of the attitude a signal-handicap for extraction from memory and for detection of records of associations in neural network models where components of vectors are represented by sign-variable pseudo-stochastic functions. It is shown, that these components essentially differ. Comparison with Hopfield-s model for calculation of size of «energy» or a vector of a condition is lead. The model offered by the author is used for an explanation of mental processes recognative and reproduction. Offered the neural network model consists of three layers - entrance, target and associative elements. Weights of communications are fixed, equiprobable and equal +1 and - 1. Communications are symmetric concerning an average associative layer. At storing factors of transfer of associative elements are measured. It is supposed, that for detection of absence of record of a vector it is expedient to use calculation of «energy». The model offered by the author is used for an explanation of mental processes узнавания and reproduction. Offered the neural network model consists of three layers - entrance, target and associative elements. Weights of communications are fixed, equiprobable and equal +1 and-1. Communications are symmetric concerning an average associative layer. At storing factors of transfer of associative elements are measured. It is supposed, that for detection of absence of record of a vector it is expedient to use calculation of «energy».